Re: How regularly is the GnuPG source code examined?

From: BRG (brg_at_nowhere.org)
Date: 09/28/05


Date: Wed, 28 Sep 2005 19:02:29 +0100

johnk55 wrote:
> Code is not verifiable! This is a conjecture, but I believe it is true.
> Verified code is only code for which the bugs have not yet been found.

This is far too general a conjecture in my view since small amounts of
code can be verified against a specification.

In fact I believe the more difficult task often ends up being that of
verifying that the context in which verification has been achieved
corresponds sufficiently well with the context in which the software
will be used to make any such verification meaningful.

> Code is used, not verified. If a lot of people use the code a lot and
> nothing goes wrong, it means either that the code is good or the bugs are
> not readily detectible.
>
> That's all.
>
> I don't believe any theoretically verifiable code exists - in spite of the
> assertions of a few academics and others touting code verification software.
> Even if theoretically possible, excellent code verification rarely happens.

Even if you are right this does not negate the value of verification
techniques since they can sometimes offer very powerful ways of
eliminating certain types of coding error that can never be conclusivley
removed by testing alone.

In particular such techniques can be used to prove that code lacks or
meets certain properties that are important even though these properties
do not constitute total correctness.

   Brian Gladman



Relevant Pages

  • Re: How regularly is the GnuPG source code examined?
    ... >>Verified code is only code for which the bugs have not yet been found. ... >There might be more to formal verification than you realize. ... It is not merely that full mechanical proofs of correctness have never ... language, such as whatever you verifier accepts as input other than ...
    (sci.crypt)
  • Re: How regularly is the GnuPG source code examined?
    ... Verified code is only code for which the bugs have not yet been found. ... excellent code verification rarely happens. ... Code is to programmers as children are to parents...you just never know what ... Let's face it, code analysis is dull, dull, dull, ...
    (sci.crypt)
  • Re: possible deadlocks?
    ... > some verification. ... They look like bugs, ... > reason why two call chains cannot happen at the same time. ...
    (freebsd-hackers)
  • Re: How regularly is the GnuPG source code examined?
    ... >Verified code is only code for which the bugs have not yet been found. ... >assertions of a few academics and others touting code verification software. ... If you are a bad programmer that is true. ... Let's face it, code analysis is dull, dull, dull, ...
    (sci.crypt)
  • RE: possible deadlocks?
    ... > My advisor Dawson Engler has written a deadlock detector, ... > some verification. ... They look like bugs, unless there is some other reason ...
    (freebsd-hackers)